International Journal of Fatigue 55 (2013) 7–21

Contents lists available at SciVerse ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Finite element simulation of fretting wear and fatigue in thin steel wires
A. Cruzado a,⇑, S.B. Leen b, M.A. Urchegui c, X. Gómez a
a
Mondragon Goi Eskola Politeknikoa, Mondragon Unibertsitatea, Loramendi 4, 20500 Arrasate–Mondragon, Spain
b
Mechanical Engineering, NUI Galway, Ireland
c
ORONA eic, Polígono Industrial Lastaola s/n, 20120 Hernani, Spain

a r t i c l e i n f o a b s t r a c t

Article history: This paper studies the effect of fretting on fatigue life reduction of thin steel wires, using the frictionally-
Received 28 November 2012 induced multiaxial contact stresses obtained from a finite element wear model, validated in previous
Received in revised form 12 April 2013 work. The fatigue life prediction model uses a critical-plane SWT approach in a 3D crossed cylinder prob-
Accepted 28 April 2013
lem. A new damage accumulation methodology for the adaptive mesh simulation, based on the cyclic
Available online 9 May 2013
material removal, has been developed. Four methods (Manson’s universal slope, Muralidharan modified
universal slopes, medians and fatigue S–N curves) for estimation of the fatigue coefficients of the wire
Keywords:
have been used. Manson’s method and medians method give lives closer to those obtained from fretting
Fretting wear
Wire ropes
wear tests in thin steel wires. The other methods are more conservative. The methodology predicts cor-
Crack nucleation rectly the life reduction of this component due to the increase of normal load (contact pressure), while it
Finite elements is not clearly predicted that an increase of the stroke reduces the life of these components as shown in the
Wear simulation experimental testing. Guidelines for developing a more robust methodology are proposed.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction centration factor produced by the wear scar and the fretting in-
duced fatigue cracks which leads to the rupture of the thin
Wire ropes combine two very useful properties: high axial steel wire. In this work it is proposed to study the effect of
strength and flexibility in bending. These properties convert wire wear-induced evolution of sub-surfaces stresses in the fretting in-
ropes into indispensable load transmission elements for many duced fatigue crack initiation.
industrial applications. For instance, wire ropes are widely used In the literature different studies about the fatigue life predic-
in cranes, mine hoisting, and lifts. These mechanical properties of tion of steel wires have been reported. Llorca and Sánchez-Gálvez
the ropes are largely dependent on their construction and the [4] presented a model to determine the fatigue limit and the fati-
properties of the wires itself, because the wires are wound into gue life to explain the results obtained experimentally in 7 mm
strands, which are then wound around a central core to form the diameter wires. This study takes into account that the fatigue frac-
wire rope, as shown in Fig. 1. The properties of the rope depend ture initiates in all cases at a surface flaw and neglect the crack ini-
on the number, size and arrangement of the wires in the strands, tiation. Beretta and Boniardi [5] proposed a suitable method for
the number and arrangement of the strands and the core type. fatigue strength prediction and quality control of wires which con-
Nevertheless, when the ropes run over the sheaves the wires sist of: tensile tests for estimating fatigue threshold of long cracks
are subjected to combined axial tensile load and bending stretch (DKth,lc), cyclic tensile tests for evaluating the theoretical fatigue
load, which leads to an oscillatory motion between the neigh- limit, determination of the extreme value of sampling and fatigue
bouring contacting wires. This phenomenon results in a fretting strength prediction using El-Haddad model. In this study wires of 1
wear damage mechanism as reported by Schrems [1] and Water- and 0.9 mm diameter were used. Petit et al. [6] presented a study
house [2]. More recently, Urchegui et al. [3] studied this effect for of fatigue crack propagation in 0.95 mm steel wires. Lambrighs
a 7  19 stranded rope, where the fretting problem is presented et al. [7] showed the influence of the inclusion parameters, such
in point contacts, typically between the outermost wires of as the inclusion size, location, composition and geometry on the fa-
strands and core or between the outermost wires of adjacent tigue properties in 0.175 mm and 0.3 mm diameter wires. Navaei
strands. As a consequence of fretting wear in point contacts, dif- et al. [8] used three damage models to predict the fatigue life of
ferent failure modes could be presented: catastrophic rupture of 1 mm diameter wires: linear damage rule (LDR), damage curve ap-
the wire due to the considerable reduction of the resistant area proach (DCA) and double linear damage rule (DLDR). After compar-
produced by the wear, fatigue life reduction due to the stress con- ing with experimental results, the DCA approach is presented as
the best method for patented steel wires. Brighenti et al. [9] ana-
⇑ Corresponding author. Tel.: +34 943 25 33 63; fax: +34 943 79 15 36. lysed the effect of the residual stresses on fatigue behaviour of a
E-mail address: [email protected] (A. Cruzado). notched wire with a surface crack subjected to alternating tension,

0142-1123/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijfatigue.2013.04.025
8 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

Nomenclature

b fatigue strength exponent Ni,l number of cycles to failure in each element for each FE
c fatigue ductility exponent fretting cycle l
dn damage of the current cycle in the surface layer n equal Nf predicted number of cycles to failure
to wk+1,n p local contact pressure
0
dn interpolated damage from the previous cycle to the cur- pav mean contact pressure
rent cycle in the surface layer n pmax maximum contact pressure
Dn damage of the first cycle equal to w1,n and accumulated Ra surface average roughness
damage in the current cycle in the surface layer n s relative slip
D ea normal strain amplitude rij ij components of local stress tensor
Dn incremental wear depth r0f fatigue ductility coefficient
Dn cycle jump rmax maximum normal stress on the critical plane
Dx stroke ru ultimate strength
E Young’s modulus ry yield strength
e0f fatigue ductility coefficient t time
f coefficient of friction T temperature
Ff friction force w Miner’s rule damage criterion
Fn normal load Wv volumetric wear
fr frequency x, y, z Cartesian coordinates
k volumetric coefficient of wear yn y coordinate of the current cycle
kl local wear coefficient Yn y coordinate of the first cycle and y coordinate of the
n layer number accumulated damage
n number of cycles
Ni number of cycles to crack initiation

showing that the crack propagation phenomenon is accelerated by and gross sliding conditions. This study showed that the crack ini-
increasing residual stresses. tiation is more likely to occur in the trailing edges of the contact.
These studies mainly show that the fatigue crack initiates from Furthermore it was reported that the crack initiation becomes
inner or surface defects presented in the steel wires. Nevertheless more difficult with increasing contact load and decreasing stroke.
in the case of wire rope systems, as has been mentioned previously, There has not yet been any methodology presented for predicting
the cracks initiate inside the wear scars, as a consequence of the both the fretting crack location and crack initiation life in steel
stress concentration factor or as a consequence of the fretting in- wires with incorporation of material removal effects. The effect
duced multiaxial stresses. Zhang et al. [10] studied the effect of of wear in some cases, such as gross sliding conditions, could be
fretting wear scars in 1 mm diameter wire obtained from 90° positive in removing cracks and lengthening life, whereas in other
crossing cylinder arrangement fretting tests that subsequently cases, such as partial or mixed slip, it could be negative by chang-
were subjected to fatigue tests. It was observed that the rupture ing the contact conditions to produce crack initiation.
of the wire was located in the deepest point of the wear scar, dem- In the recent years many researchers have introduced the mul-
onstrating that the fatigue life was inversely proportional to the tiaxial fatigue models to establish life predictions in fretting prob-
fretting wear depth. The main reason of this effect is the stress con- lems. Szolwinski and Farris [14] were the first researches that
centration at this point and the following crack initiation, propaga-
tion and fracture of fretted wire. Following these tests and
including fretting wear scars for different crossing angles, Wang
et al. [11] proposed an equation to investigate the effect of fretting
wear on fatigue fracture properties of steel wires assuming that the
fretted scar (fretting wear depth) on the wire surface was the ini-
tial pre-crack. To this end the Paris equation and an equation
which represent the evolution of dimensionless stress intensity
factor for crack growth in pure tension was proposed. The dimen-
sionless stress intensity factor was obtained from the assumption
that the crack assumes a semicircular shape from very early and
tends to flatten. Recently, in order to analyse the effect of fret-
ting-fatigue in 1 mm diameter wire, Wang et al. [12] have pro-
posed a new fretting fatigue test rig for steel wires. All the tests
were performed under mixed fretting regime conditions. It was
concluded that an increase of displacement amplitude at each cyc-
lic strain level increases the tangential force and the relative slip
range at the stabilized stage, accelerating the fretting damage of
contact surface. Based on this experimental configuration Wang
et al. [13] performed a numerical FE analysis employing the multi-
axial fatigue criteria of Fatemi–Socie and Smith–Watson–Topper to
analyse the effect of fretting parameters on the crack initiation
location in the initial fretting fatigue stage in both partial sliding Fig. 1. Schematic of a wire rope composed of different strands of wound steel wires.
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 9

introduced the critical plane methods with the Smith–Watson– from fretting wear tests of thin steel wires, giving in the end guide-
Topper (SWT) parameter to predict fretting fatigue life from fret- lines for the construction of a more robust tool in future works.
ting induced stresses. Araujo and Nowel [15] demonstrated that
the SWT and the Fatemi–Socie (FS) multiaxial fatigue models com- 2. Experimental details
bined with analytical approach such us Mushkelishvili potential
theory to determine the two dimensional contact stress field in cyl- 2.1. Material
inder on flat configurations, resulted in over conservative predic-
tions at smaller contacts due to the presence of high stress The wires used for the experiments are wires which are usually
gradients. So they proposed two averaging methods, which com- fabricated into a 7  19 rope (7 strands with 19 individual wires in
bined with the multiaxial fatigue criteria, resulted in a less conser- each one) as presented in Fig. 1. For the tests, wires with a diame-
vative prediction. Fouvry et al. [16] also used an averaged stress ter of 0.45 mm, were used. They are cold-drawn from eutectoid
approach, due to the very sharp stress gradient imposed below carbon steel with 0.8% C with a hardness of 659 ± 81 HV0.05. The
the surface in a cylinder on plane configuration. It was proposed tensile properties obtained from the measured stress–strain curve
the introduction of a representative damage volume with the shown in Fig. 2, were an ultimate strength of 3000 MPa and an off-
Crossland’s multiaxial fatigue criterion for predicting the crack set 0.2% yield strength of 2688 Mpa. The surface average roughness
nucleation under stabilized partial slip conditions. (Ra) of these wires along its axis was 0.35 and 0.70 lm in the per-
Sum et al. [17] developed a finite element based implementa- pendicular direction. These values were obtained from the mea-
tion of the critical plane SWT fatigue life method. It was shown surements carried out on an unworn wire using a confocal
that this methodology capture the well known contact size effect imaging profiler (Pll, Sensofar). The tensile properties and the
of the cylinder plane configuration shown previously by Araujo Vickers hardness are summarized in Table 1.
and Nowell [15]. Madge et al. [18] proposed a finite element based
method to predict and characterise the role of fretting wear on 2.2. Tribological testing
fretting fatigue for a rounded punch-on-flat Ti–6Al–4V under par-
tial slip and gross sliding conditions. This model combined a finite An extended description of the experimental testing and results
element wear simulation model with the critical plane SWT fatigue used to study the fretting induced cracks in thin steel wires is re-
model. They concluded that the prediction of fretting wear was a ported in [20]. In this section, the test rig is briefly outlined and
critical aspect for predicting fretting fatigue, which cannot be pre- the conditions that were selected for the analysis of the subsurface
dicted by models that do not include the effects of material re- stresses and the crack initiation life prediction using the FE damage
moval. Recently Zhang et al. [19] used a similar model in a round approach proposed in this paper.
on flat and rounded punch on flat configuration for the wear evo- Fretting wear tests between thin steel wires were carried out on
lution and crack nucleation predictions. It was concluded that if a specific tribometer for small sliding amplitudes (10–200 lm) and
the modelling of wear was not taken into account the fatigue dam- small loads (0.1–10 N). The tribometer consists of a top arm with
age prediction could be under-conservative in the partial slip re- an integrated load cell for the measurement of the friction force
gime (PSR) and over conservative in the gross slip regime (GSR). and a bottom vibrating horizontal table. The movement of this ta-
Based on the different studies reported in the literature, firstly ble is given by a frequency-controlled DC motor with an eccentric
the analysis of the fatigue life prediction for wire ropes requires adjustment. Special holders were designed for the correct fixing of
investigations of the wear contact problem originated changes in the wires. In these holders the wires represent a toroid with a cur-
the contact region, because the reciprocal sliding between the vature radius of 10 mm. The bottom holder is fixed to the vibrating
wires, which generates considerable wear, can modify the wire horizontal table in the tribometer, with the wire axis parallel to the
shapes in contact permanently, both during tension and bending direction of motion. The upper holder (stationary sample), is fixed
of the rope, that implies the continuum change of the subsurface at the end of a horizontal arm forming the desired crossing angle
stresses. Secondly in order to acquire a better understanding of between its axis and the bottom wire axis. During each test the real
the fatigue behaviour of ropes, the severe fatigue degradations that stroke (Dxreal), friction force (Ff) and total linear wear (Wl,tot) were
occur between contacting wires has to be implemented using mul- measured. Relative humidity (RH) and temperature (T) were kept
tiaxial fatigue models. constant during the test. All measured quantities were recorded
In previous work [20] the influence of fretting wear in on-line multiple times. After each test the frictional and wear
0.45 mm diameter thin steel wires under different loading condi- quantities used in the FE wear model were obtained.
tions was analysed. In these tests, it was detected that under
high normal loads and strokes cracks initiate at a certain number
of cycles. More recently with the aim to predict fretting wear
scars and reduce the time consumed and the associated cost that
requires the experimental data, an optimized FE wear simulation
approach for thin steel wires was developed and validated [21],
based on the fretting wear tests carried out in the previous work.
As a consequence of the extremely complex stress range ap-
peared in this type of components, due to the complex structure
and behaviour shown by the wire rope systems, the present paper
is the first step to develop a FE tool for predicting fretting induced
cracks in wire rope systems. In this work it is proposed to study the
fretting wear induced fatigue life reduction of the wires, using the
frictionally induced multiaxial contact stresses obtained from a
FEM wear model developed previously. To this end the multiaxial
critical plane SWT approach with a new damage accumulative
model, which takes into account the material removal in the accu-
mulated damage, has been developed. The results obtained from Fig. 2. Experimentally obtained engineering stress–strain curve for 0.45 mm
this tool have been compared with the experimental data obtained diameter wire.
10 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

Table 1
Mechanical properties of the 0.45 mm diameter wire.

E (GPa) ru (MPa) ry (MPa) % RA HV0.05

Wire 200 3000 2688 33 659 ± 81

The tests were carried out with a 90° crossing angle configura-
tion for a range of mean contact pressures that goes from
2460 MPa to 3933 MPa, two strokes 65 lm and 130 lm in a range
of cycles from 20,000 to 200,000. The specific details of these tests
are shown in Table 2. After the tests the contact surfaces where
analysed by optical microscopy in order to find crack and no crack
conditions. Only fretting cracks were found in the bottom speci-
men, so the FE analysis will be focused on this specimen.

3. FEM methodology

3.1. FE model

Fig. 3. FE model and mesh details of the 90° crossed cylinder fretting wear test.
The FE model used in this study is the same that has been val-
idated in a previous work [21], for the simulation fretting wear in
capture precisely the variation of contact pressures and geometry.
thin steel wires under different loads conditions. The model con-
The mesh is composed of square shape eight node linear brick ele-
sists of two cylinders with a diameter of 0.45 mm contacting to-
ments. The size of the meshes in the contact zone is defined follow-
gether at 90° crossing angle as shown in Fig. 3. Moreover the
ing the optimized mesh criterion to reduce the computational time
curvature radius that acquires the wire in the fixation support
proposed for fretting wear scars modelling [21]. Taking into ac-
(10 mm) is included.
count that, an optimum mesh should be in the range from 3% to
The load and the boundary conditions are applied in the free sur-
4% of the final longitudinal wear width and that the dimensions
faces of the model through a surface-based coupling, which couples
of the wear scars analysed in this study are in the range from
the motion of all free surfaces to the motion of a reference node.
250 lm to 310 lm, a 10 lm  10 lm  7.5 lm mesh size which
Moreover a kinematic coupling is used, so that the rigid body mo-
corresponds to 4% of the smallest wear scar dimension is chosen.
tion is defined by the reference node. It allows in the upper wire,
As reported in [21] smaller mesh sizes increase the computational
where a normal force is applied in the reference point, only vertical
time without any improvement in the final result and greater mesh
movement and in the bottom wire, where an alternative displace-
sizes give a poor discretization of the wear scar.
ment condition is imposed, only horizontal movement. This is be-
The contact surface interaction between both cylinders is de-
cause the wire follows the rigid body movement of the holder in
fined via the finite sliding contact pair algorithm, which uses the
the testing machine. Elastic isotropic material model was specified
master–slave algorithm to enforce the contact constraints. In this
in both cylinders, in which a standard values for steel (Young’s
study the top surface is chosen as slave surface and the bottom
modulus of 210 GPa and Poisson’s ratio of 0.33) have been used.
one as master. The Lagrange multiplier contact algorithm was used
Although in this study contact pressures greater than the yield
to ensure the exact stick constraint when the shear stress is less
strength of the material which produce plastic deformation were
than the critical value according to the Coulomb friction law. A
used, an elastic plastic model was not defined. On the one hand
coefficient of friction f of 0.7 is chosen according to the experimen-
plastic deformations only are given in the first cycles. The wear pro-
tal tests carried out in thin steel wires.
duces a rapid increase of contact area which leads into a drastic de-
crease of contact pressure, giving negligible effect of plastic
deformation along the fretting tests. On the other hand the intro- 3.2. Wear modelling
duction of an elastic plastic model produces convergence problems
in the contact resolution and increases the computational time. The specific method used in this study for the wear simulation
The meshes are carried out according to a partition methodol- of thin steel wires, which was validated satisfactorily under differ-
ogy, in which a more refined mesh is used in the contact zone, to ent range of loads and slip amplitudes, is described in detail in the

Table 2
Fretting test conditions for the study of crack initiation process in 90° crossed cylinders steel wires.

Parameter Symbol Value Tribosystem

Normal load (N) Fn 1, 2, 3, 4
Mean contact pressure (MPa) pm 2460; 3100; 3600; 3933
Maximum contact pressure (MPa) pmax 3690; 4650; 5400; 5900
Stroke (lm) Dx 65 ± 5 (2 N, 3 N, 4 N)/130 ± 5 (1 N, 2 N)
Frequency (Hz) fr 10
Number of cycles (103) N 20; 50; 100; 200
Lubricant None
Temperature (°C) T 25 ± 1
Relative humidity (%) RH 50 ± 5
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 11

work reported by Cruzado et al. [21]. The wear model used for the 3.3. Damage modelling
fretting wear simulation is the Archard’s modified equation
proposed by McColl et al. [22]. This equation assumes that fretting Fretting wear in thin steel wires leads to a frictionally induced
wear can be evaluated by applying Archard’s equation to local con- multiaxial contact stress distribution, which continuously changes
tact conditions along the entire contact interface. The local wear due to the wear-induced evolution of the subsurface stress. This
model is described by: condition with the combination of high loads and strokes leads
to fatigue cracks in the contact surface. Moreover during these
Dhðx; tÞ ¼ kl  pðx; tÞ  sðx; tÞ ð1Þ tests, the fretting-induced multiaxial stress field is combined with
the high tensile stress produced as a consequence of the bending
that the wire suffers in the fixation holder. The greater influence
where Dh(x, t), p(x, t) and s(x, t) are the incremental wear depth, the
of the tensile stresses, as will be discussed in Section 4.2 leads to
contact pressure and the relative slip for a specific point x (node) at
mode I cracks.
specific time t, respectively. On the other hand kl represents the lo-
With the purpose to study the multiaxial fatigue phenomenon
cal wear coefficient. This coefficient in principle could not be the
presented in thin steel wires, the critical plane Smith–Watson–
same as the volumetric coefficient of wear k, because the last one
Topper (SWT) multiaxial fatigue criterion, generally used in cases
represents an average value across the complete wear scar. Never-
involving tensile cracking, has been implemented. This criterion
theless some authors like McColl et al. [22] or Fouvry et al. [23] have
has been applied satisfactorily by many researchers for the predic-
justified the close relation between the local wear coefficient kl and
tion of crack nucleation life and location in fretting problems, both
the volumetric coefficient of wear k. In this work the volumetric
in fretting fatigue loading conditions [14,15,17,26] and in fretting
coefficients of wear k corresponding to each specimen and test con-
wear loading conditions [19,27,28]. The SWT parameter predicts
ditions, which are shown in Table 3, have been used. It has to be
that the crack initiation and growth occurs on a certain plain,
pointed out that the wear simulations presented in this study will
where the normal stress and strain in that plane is maximum.
be carried till 50,000 cycles.
The parameter is shown in Eq. (3), where rmax is the maximum
Taking into account Eq. (1), one fretting cycle is divided in small
normal stress in the critical plane and Dea is the normal strain
slip increments; it implies that the FE wear simulation is the prod-
amplitude in the same plane, E is the Young modulus, r0f and b
uct between these increments and the total number of cycles. As a
are the Basquin strength coefficient and exponents corresponding
consequence, high computational time is needed for the simulation
to the high cycle fatigue (HCF), e0f and c are the Coffin–Manson fa-
of one fretting test. In this point the cycle jumping technique pro-
tigue ductility constants corresponding to low cycle fatigue (LCF)
posed by different authors like McColl et al. [22] and Mary et al.
and Ni represents the number of cycles needed for the crack
[24] is used, where it is assumed that the contact pressure and
initiation.
the slip distribution remain constant in each finite element analy-
sis over Dn cycles. In this way Eq. (1) is multiplied by Dn to obtain
Eq. (2), such that the wear corresponding to Dn cycles can be sim- r02f
ulated in one FE fretting cycle. In the study carried out by Cruzado SWT ¼ rmax  Dea ¼ ð2Ni Þ2b þ r0f e0f ð2Ni Þbþc ð3Þ
E
et al. [21] it was reported that an optimum simulation without any
The implementation of this method in the FE model presented
distortions in profile should be carried out with 100 fretting FE cy-
previously has been carried out according to the methodology pro-
cles, 40 increments/fretting FE cycle and the corresponding Dn. Dn
posed by Sum et al. [17] and Das and Sivakumar [29] for 3D prob-
is obtained dividing the desired number of cycles n by the 100 fret-
lems. This methodology consists of the three-dimensional
ting FE cycles that compose one FE simulation, e.g. for 50,000 num-
transformation (Mohr’s circle) of the stress and strains of each ele-
ber of cycles a Dn of 500 is used.
ment centroid at intervals of 5° over 180°, analyzing all the planes
that can be found in a semi-sphere. So in each centroid are re-
Dhðx; tÞ ¼ Dn  kl  pðx; tÞ  sðx; tÞ ð2Þ corded 1296 planes for each element and for each increment that
The implementation of the wear model has been carried out form one fretting cycle. From this data the maximum normal stress
using the ABAQUS user subroutine UMESHMOTION [25], in which rmax and the normal strain amplitude Dea are recorded in each
the spatial adjustment of the contact nodes is done within an adap- plane of each element, where Dea represents the difference be-
tive framework. Nevertheless this subroutine only gives the con- tween the maximum and the minimum value of strain normal to
tact results for one of the contacting surfaces (slave surface), i.e., the candidate plane over the entire fretting cycle. Finally the
so the transition of the contact pressures and slips from the slave SWT values are obtained for each candidate plane and each ele-
surface to the master surface is done according to the nearest inter- ment, which are then used to obtain the maximum SWT value in
polation method, since this method is computationally most effi- the critical plane of each element. With the maximum SWT value
cient [21]. and solving Eq. (3), the numbers of cycles to failure for each fret-
ting cycle are obtained.
Nevertheless, due to the material removal effect, the stresses and
Table 3 strain change from cycle to cycle. To capture this phenomenon Eq. (3)
Coefficients of wear obtained from fretting wear tests under different loads–stroke
is used to predict the number of cycles to failure Ni,l in each element
conditions.
for each FE fretting cycle l, where the damage corresponding to each
Coefficient of wear k Coefficient of wear k FE fretting cycle l is accumulated using the Miner–Palmgren linear fa-
(108 mm3/Nmm) (108 mm3/Nmm)
tigue accumulation damage rule shown in Eq. (4). Thus the crack
Top–bottom Top–bottom
nucleation is occurred when the total accumulated damage w in a
2 N–65 lm– 50,000 cycles 2.11–1.57 1 N–130 lm– material point reaches 1. In this study, as the wear simulation prob-
50,000 cycles
2.32–2.34
lem is computationally costly, a cycle jump technique was used for
3 N–65 lm– 50,000 cycles 2.10–1.91 2 N–130 lm– the wear modelling optimization. Therefore, the damage accumula-
50,000 cycles tion model is expressed as Eq. (4), where Ni,l is the critical-plane
2.32–2.34 SWT predicted number of cycles for crack initiation at fretting FE cycle
4 N–65 lm– 50,000 cycles 2.32–2.34
l, n is the desired number of cycles simulated in the FE fretting wear
model and Dn is the cycle jump value. The ratio between n and Dn
12 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

represents the 100 fretting FE cycles of which one FE simulation is

composed, as mentioned in Section 3.2.
n
X
Dn
Dn
w¼ ð4Þ
l¼1
Ni;l

On the other hand, as a consequence of the adaptive meshing

used in the FE wear simulation approach to update the mesh due
to the material removal, the damage cannot be accumulated in
the same centroidal point from cycle to cycle. The reason is be-
cause the centroidal position of each element is changing from cy-
cle to cycle, so the damage accumulated in the centroid position of
a previous wear cycle could not be the same in the new centroidal
position of the current cycle. The previously accumulated damage
in this point should be less, such as the damage goes decreasing
from the surface to the inner of the material. So the effect of mate-
rial removal in the damage eradication has to be considered.
To solve this problem Madge et al. [18] proposed to create a
material point mesh (MPM) as the global reference model for dam-
age accumulation. In this method the nodes of the MPM have fixed
coordinates through the analysis, so the cyclic damage calculated
at the centroid of each element was linearly interpolated back to
the MPM for accumulation, and the nodes corresponding to the re-
moved material do not accumulate any fatigue damage.
In this study, due to the high amount of wear produced by the
fretted thin steel wires a new damage accumulation model based
on the previous and on the current location position of the element
centroid due to wear has been developed. This approach leads to
update the damage in the new centroidal position after the FE wear
simulation cycle. The mesh used for the wear modelling in the
adaptive meshing framework is shown in Fig 4. Taking into account
Fig. 5. Flow chart of the accumulative damage model due to the effect of the
that the crack initiation is produced close to the surface layers and
material removal.
due to the great amount of data that is collected in a 3D problem,
only the three first layers of the centroid points will be used in this
new methodology. Dn is the cycle jump value and Ni,1,n is the critical-plane SWT
A flowchart which summarises the proposed process for deter- predicted number of cycles for crack initiation at the first fret-
mining the accumulation of damage due to the wear removal is ting FE cycle of each element in the first three layers n.
shown in Fig. 5.
So the following methodology for the application the accumula- Dn
tive damage rule due to wear is outlined here: Dn ¼ w1;n ¼ ð5Þ
Ni;1;n
1. Cyclic damage Dn due to the first fretting cycle is calculated at 2. As a consequence of the material removal, the damage obtained
the centroid point location Yn of each element in the first three in the centroidal points Yn of the first cycle will not be main-
layers n (Fig. 4b left). Dn is calculated according to Eq. (5), where tained in the new centroidal position location yn reached after
the second cycle due to wear (Fig. 4b right): so two linear

(a)

(b)
Fig. 4. FE mesh in the adaptative meshing framework and each centroid points: (a) 3D FE mesh with a cut of one of the longitudinal plans and (b) FE mesh with the centroid
points before and after the wear process.
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 13

(a) (b)

d1

Damage
Damage
D1 D1
D1 (Y )

d2
D2 D2 (Y ) D2
d3
D3 D3
Y1 Y2 Y3 Y1 y1 Y2 y2 Y3 y 3
y coord [mm] y coord [mm]

d1 + d1'
(c) (d) D 1

D1 (Y )

Damage
Damage

D1
d1'
D2 d 2 + d 2'
D2 (Y )
d 2'
D2 d3 + d3'
d3' D3
D3
Y1 y1 Y2 y2 Y3 y 3 Y1 = y1 Y2 = y2 Y3 = y 3
y coord [mm] y coord [mm]

Fig. 6. Damage accumulation process: (a) damage of the first cycle (circle) in the first three layers and the two linear equations which represent the evolution of the damage
between the layers, (b) damage in the second FE cycle due to wear (square), (c) damage of the first cycle interpolated back to the new centroid position of the second cycle (x),
and (d) accumulated damage in the second cycle (triangle) and the new linear equations (graphic b (square) plus graphic c (x)).

equations as shown in Eqs. (6) and (7), which represent the 5. In the new centroid position (Eq. (12)) of each element, as it is
0
damage evolution between the first and the second layer and shown in Fig. 6d, the interpolated damage dn due to the first
the second and the third layer, are constructed. The graphic rep- cycle is accumulated with the damage of the second cycle dn
resentation is shown in Fig. 6a. (Eq. (13)).
D2  D1 Y n ¼ yn ð12Þ
D1 ðYÞ ¼ ðY  Y 1 Þ þ D1 ð6Þ
Y2  Y1

0
D3  D2 Dn ¼ dn þ dn ð13Þ
D2 ðYÞ ¼ ðY  Y 2 Þ þ D2 ð7Þ
Y3  Y2
6. Two new linear equations (Eqs. (6) and (7)) are constructed for
3. Cyclic damage dn due to the second fretting cycle is calculated the new centroid position due to wear (Eq. (12)), and the new
at the new centroid point location yn of each element in the first damage values (Eq. (13)), as represented in Fig 6d.
three layers as it is shown in Fig. 6b. dn is calculated according 7. This process is repeated as many times as number of cycles
to Eq. (8), where Dn is the cycle jump value and Ni,l+1,n is the needed to reach the damage failure, when Eq. (13) reaches 1.
critical-plane SWT predicted number of cycles for crack initia-
tion at the second fretting FE cycle of each element in the first In order to reduce the postprocessing time required by this
three layers n. damage accumulation process, and knowing that the relocation
of the new centroidal position in the third layer after each wear
Dn and remeshing cycle is very small, an extrapolation of the third
dn ¼ wlþ1;n ¼ ð8Þ
Ni;lþ1;n layer damage using the equation corresponding to the second
4. The damage of the first fretting cycle is interpolated back to the and third layer Eq. (7) was carried out as shown in Fig. 6c. There-
new position at the centroidal point’s location yn of the second fore, taking into account that Y3 ’ y3 the error committed in the
cycle, as presented in Eqs. (9)–(11) and in Fig. 6c. damage reduction calculation of the previous cycle in the current
cycle is very small. This error will be accumulated from cycle to cy-
0 D2  D1 cle in the construction of the new linear equation of both layers
d1 ¼ ðy  Y 1 Þ þ D1 ð9Þ
Y2  Y1 1 Eqs. (6) and (7); nevertheless as the damage of the second and
third layer is very small comparing with the damage of the first
layer, this error is negligible.
0 D3  D2
d2 ¼ ðy  Y 2 Þ þ D2 ð10Þ
Y3  Y2 2
3.4. Material fatigue constants

In the absence of fatigue properties of the wire studied in this

0 D3  D2
d3 ¼ ðy  Y 2 Þ þ D2 ð11Þ work, different approaches, for obtaining the fatigue strength coef-
Y3  Y2 3
ficients and the fatigue ductility coefficients needed for the life
14 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

Table 4 cients obtained from the S–N curves reported by Beretta and Mat-
Estimation methods of fatigue coefficients from monotonic tensile tests [31–33]. teazzi [34] in a 1.97 mm diameter wire.
Parameter Universal slopes Modified universal slopes Medians (steels)
0
r f 1.9 ru 0.623 r 0:823
u E 0.168 1.5 ru 4. Results and discussions
b 0.12 0.09 0.09
e0f 0.76 e0:6 0.0196 e0:6 (ru/E)0.53 0.45
f f
The FE analysis presented in this section corresponds to the
c 0.6 0.56 0.59
fretting wear tests carried out on a 0.45 mm diameter pre-stressed
wire, as a consequence of the bending that was applied in the wire
for the correct fixation in the holder. In order to analyse the fretting
Table 5 wear frictionally induced crack nucleation, firstly a study of the
Mechanical properties of Beretta’s 1.97 mm diameter wire [34]. subsurface stress distribution in a not pre-stressed steel wire has
ru (MPa) ry (MPa) % RA %E HV been carried out. Secondly the effect of the bending imposed pre-
stresses in the fretting-induced subsurface stresses has been ana-
Beretta 1855 1540 58 1.61 470 ± 30
lysed. Thirdly the results obtained from combined FE wear model
and the critical plane SWT–damage approach proposed in this pa-
per has been discussed and compared with the results obtained
prediction according to the SWT model (Eq. (3)), are discussed in experimentally. Finally different strategies for the improvement
this section. of the developed methodology for predicting fretting wear cracks
The first approach is based on the methods proposed by differ- in thin steel wires have been proposed.
ent authors for estimating the fatigue coefficients based only on
the available monotonic tensile test data. This methodology has
been used previously by Del Llano-Vizcaya et al. [30] for predicting 4.1. Subsurface stress distribution analysis
the fatigue lives in helical compression springs which are made by
high carbon steel wires, using different multiaxial fatigue criteria. The wear-induced evolution of sub-surface stresses which di-
In Table 4 are shown the different estimation methods proposed rectly controls the crack initiation in the bottom cylinder after dif-
in this study. The first one corresponds to the Manson’s universal ferent numbers of cycles and 1 N–130 lm is discussed in the
slope method [31], which takes into account the ultimate strength following paragraphs. A three dimensional local axis coordinate
of the material ru and the true fracture stain ef which is calculated system with the local 1 direction parallel to the direction of the
using the reduction in area RA ef = ln (1/1  (RA/100)). The second movement (x direction in Fig. 3); the local 3 direction normal to
one is a modified universal slopes method proposed by Muralidh- the contact surface (z direction in Fig. 3), and the local 2 direction
aran and Manson [32], in which both fatigue exponents are in- making a right hand set with the other two has been used. The sub-
creased and the ru/E term is introduced for the calculation of surface stresses have been plotted in the local 1 direction parallel
both fatigue coefficients. The last method named medians method to the direction of the movement, in the central path y = 0 of the
for steel is introduced by Meggiolaro and Castro [33], which com- contact at a depth of 3.5 lm, which corresponds to the first row
bines good average life predictions with one of the lowest standard of centroidal values of the lower specimen. This path matches with
deviations for a higher range of different steels. the maximum SWT location that will be presented in Section 4.3.
The second approach is based on obtaining the fatigue proper- Furthermore all subsurface stresses plotted in this paper are
ties from an eutectoid high strength steel wire reported in the lit- belonging to the left hand side motion of the lower wire. It means
erature. So, the Basquin’s (HCF) coefficients were obtained from that the contact area is moving to the right hand side and thus the
the R = 0.1 uniaxial fatigue tests reported by Beretta and Matteazzi sliding between the upper (static) and the lower wires is produced
[34] in a 1.97 mm diameter wire with an interlamellar spacing of in this direction. At the mirror, these figures represent the opposite
about 0.1 lm. The mechanical properties of this wire are shown direction motion of the lower wire.
in Table 5 and are in accordance with those ones reported by Ver- In Fig. 7 the r11 and r13 stresses after 20,000 fretting wear cy-
poest et al. [35] for 1.99 mm diameter wire. To obtain the r0f and b cles and different displacement amplitude locations are shown. It
in fully alternated tension–compression fatigue tests R = 1 the can be seen from these plots that there is an increase of the peak
standard Goodman relationship was used to correct for the mean stress value from the centre position to the end position of the dis-
stress effect. In the absence of low cycle strain controlled fatigue placement amplitude, with the greater value in this last position.
data, e0f and c were obtained applying the monotonic tensile stress This is attributed to the shape acquired by the wear scar, composed
properties in the Manson’s universal slope method presented by a uniform surface geometry that goes from the centre to the
previously. near edge of the wear scar and with inclined flanks in the edge
A summary of the fatigue coefficients obtained from the mono- of the wear scar.
tonic tensile test properties of the studied steel wire shown in Ta- Fig. 8 shows the multiaxial subsurface stress distribution for
ble 1 and for the different methods proposed in Table 4 are shown different numbers of cycles at a displacement amplitude position
in Table 6. Moreover in this table are included the fatigue coeffi- of 65 lm which corresponds to the maximum peak value pre-
sented in Fig. 7. It can be seen that there is a drastic reduction in
the peak stress value before and after wear, as a consequence of
the decrease at the peak contact pressure in the first few hundred
Table 6
of cycles as reported in [21]. On the other hand, while a tensile
A summary of the fatigue coefficients obtained from the different methods. stress parallel to the surface promotes crack nucleation, a compres-
sive stress retards it. Taking into account this affirmation, the r11
Parameter Universal Modified universal Medians Beretta
slopes slopes (steels)
axial stress in the direction of sliding and the r13 shear stress are
the key components in the crack nucleation process, as r22 mainly
r0f 5700 3521 4500 8000
presents compressive stresses and r12 is the shear stress appeared
b 0.12 0.09 0.09 0.225
e0f 0.441 0.1053 0.45 0.697
in the transversal direction to the movement, located in the x coor-
c 0.6 0.56 0.59 0.6 dinates of the maximum r11 and r13 value. The value of r12 in the
longitudinal path is 0 MPa. Moreover while r13 only gives positive
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 15

(a) 1000
(b) 1000
0 0

σ11 [MPa]

σ13 [MPa]
-1000 -1000
0 μm 0 μm

-2000 19.5 μm -2000 19.5μm

39 μm 39 μm
65 μm 65 μm
-3000 -3000
-0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2
x coord [mm] x coord [mm]

Fig. 7. Subsurface stress distribution for 1 N–130 lm, 20,000 cycles and different amplitude displacement locations: (a) stress distribution r11 and (b) stress distribution r13.

(a) 1000 (b) 1000

0 0
σ11 [MPa]

σ22 [MPa]
-1000 -1000
Sliding direction

-2000 Zer o cycle -2000 Zer o cycle

20000 cycles 20000 cycles
50000 cycles 50000 cycles
-3000 -3000
-0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2
x coord [mm] x coord [mm]

3000
(c) Zer o cycle
(d) 600

20000 cycles 400

2000
50000 cycles
200
σ13 [MPa]

σ12 [MPa]

1000 0

-200
0 Zer o cycle
-400 20000 cycles
50000 cycles
-1000 -600
-0.2 -0.1 0 0.1 0.2 -0.1 -0.05 0 0.05 0.1
x coord [mm] y coord [mm]

Fig. 8. Multiaxial subsurface stress distribution in the bottom cylinder for 1 N–130 lm and different number of cycles in the amplitude displacement position of 65 lm: (a)
stress distribution r11, (b) stress distribution r22, (c) shear stress distribution r13 and (d) shear stress distribution r12.

stresses, r11 gives both positive and negative. It can be seen in these components present the smallest influence in the crack initi-
Fig. 8a that while the leading edge of the contact (the front side ation as has been shown previously and the coefficient of friction
in the sliding direction) is subjected to compressive stresses, the has low influence in this two stresses. In this study these compo-
trailing edge of the contact (the back side in the sliding direction) nents are governed by the contacting load, such that r22 represents
is subjected to a tensile stresses. Furthermore, due to the increase the compressive stress and r12 the shear stress produced by the
of the length of the wear scar, a translation of the peak stresses in normal load. If a rotation or a tensional moment in the x axis
the sliding direction is produced. This effect is smaller as the num- was introduced, r12 would have greater influence.
ber of cycles increases. A reduction of the coefficient of friction implies a reduction of
When analyzing the values of r11, r13 and r12 (r22 is not con- the subsurface stresses for both the tensile axial stress r11 and
sidered because only gives compressive stresses), it can be derived the shear stress r13. It means that reducing the coefficient of fric-
that the shear stress, which corresponds to the shear stress in the tion can increase the crack nucleation life. As has been mentioned
sliding direction, is the most critical. This tendency is maintained in Section 3.1, in this study a friction coefficient of 0.7, which is in
with the number of cycles. As an example it can be seen that while accordance with the coefficient of friction obtained experimentally
for 50,000 numbers of cycles r13 = 455 MPa, r11 = 181 MPa and in the stable period [20] has been used.
r12 = 100 MPa.
The effect of coefficient of friction in the previously shown key 4.2. Subsurface stress distribution in a pre-stressed wire
components on the crack nucleation, the axial stress r11 and the
shear stress r13 for 1 N–130 lm before wear is illustrated in In the previous paragraphs the analysis of the fretting induced
Fig. 9. The r22 and the r12 are not plotted in this study, because subsurface multiaxial stress distribution has been described. Nev-
16 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

1000 2000
0.2
500 1500
0.45
0 1000 0.7

σ11 [MPa]

σ13 [MPa]
-500 500

-1000 0.2 0

-1500 0.45 -500

0.7
-2000 -1000
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.15 -0.1 -0.05 0 0.05 0.1 0.15
x coord [mm] x coord [mm]

Fig. 9. The axial subsurface distribution for 1 N–130 lm and different coefficient of friction: (a) stress distribution r11 and (b) shear stress distribution r13.

ertheless, the curvature used to correctly fix the wire in the holder sile pressure of 2000 MPa in the extremes of the wire, has been
produces a bending deformation in the wire, which leads to tensile introduced in the bottom cylinder of the fretting wear model.
stresses near the contact surface. Fig. 11 shows the axial stress r11 and the shear stress r13 for
To analyse the stress field produced in the wire as a conse- 1 N–130 lm and 50,000 number of cycles, with and without pre-
quence of the bending, Fig. 10 shows the FE model used for mod- stresses. The r22 and r12 stresses are not considered in this analy-
elling the wire fixation procedure in the holder. While the wire sis because the tensile stress introduced as a consequence of the
supporting holder and the two fixation washers have been mod- bending has no influence on these two stresses.
elled as a 3D analytical rigid body, the wire has been modelled The effect of the pre-stress is significant in the r11 axial stress,
according to an elastic plastic model with isotropic hardening. where an increase of the tensile stresses along the entire contact
The elastic plastic model has been introduced using the measured region is shown. Moreover there is a decrease of the tensile stres-
stress–strain curve shown in Fig. 2. The mesh is composed of eight ses in the front side of the contact as a consequence of the com-
nodes linear brick elements. The contact surface interactions be- pressive stresses produced by the leading edge of the contact, as
tween the rigid surfaces and the wire are defined via the finite slid- has been discussed previously. On the other hand the r13 shear
ing contact pair algorithm. With the aim to simplify the problem stress presents a less significant increase. This suggests that a de-
no frictional contact has been considered. crease of the wire pre-stresses, e.g. increasing the bending radius,
Fig. 10 shows the stress produced in the axial direction of the can increase the crack nucleation life. Moreover it can be derived
wire r11 which is in accordance with the axis coordinate system that while in a non pre-stressed wire the shear stress r13 is the
defined in the previous section. This stress, in combination with most critical, in the pre-stressed wire the tensile traction r11 be-
fretting induced multiaxial stresses, will affect the crack nucleation comes the most critical value in the crack initiation, leading to
process. It can be seen that in the region corresponding to the con- cracks in mode I.
tact surface in fretting tests, the model predicts a compressive
stress in the bottom side of the wire and a tensile stress in the 4.3. Fatigue crack nucleation analysis
top side of the wire which corresponds to the contact surface.
The results show that there is a tensile stress r11  2000 MPa near Firstly, the analysis of the crack nucleation prediction due to
the contact surface. On the other hand the stresses r22 and r33 in fretting wear in thin steel wires has been carried out according
the contact region are less than 5% of the stress obtained in r11, to the proposed FE methodology. Secondly, the results have been
so they will be neglected in the study of the combined fretting compared with the experimental data obtained from fretting wear
and bending. tests.
To analyse the effect of the pre-stresses in the fretting wear Fig. 12 illustrates the critical plane SWT crack nucleation pre-
analysis, an initial predefined state, obtained after applying a ten- dictions plots for the case of 1 N–130 lm pre-stressed wire at

Fig. 10. Elastic–plastic finite element analysis of the wire fixation r11 (MPa) axial stresses.
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 17

3000 3000

2000 2000

1000 1000

σ13 [MPa]
σ 11 [MPa] 0 0

-1000 -1000

0 MP a 0 MP a
-2000 -2000
2000 MP a 2000 MP a

-3000 -3000
-0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2
x coord [mm] x coord [mm]

Fig. 11. The axial subsurface stress distribution for 1 N–130 lm, 50,000 cycles and different pre-stresses in the amplitude displacement position of 65 lm: (a) stress
distribution r11 and (b) shear stress distribution r13.

2000 MPa. Fig. 12a shows the maximum SWT critical plane value graphics, the no wear case and the wear case, it is found that
along the axial direction of the bottom wire (sliding direction). the path corresponding to the maximum SWT critical plane value
The results show that there is a drastic decrease of the SWT peak along the direction of the movement, is located in the central
value (Fig. 12b) in the first thousand of cycles and then, this value path y = 0.
stabilizes and decreases at a lower rate. This effect is attributed to Fig. 13 shows the effect of the bending induced pre-stresses in
the material removal produced by the wear, which causes stress the SWT peak value. In both cases the pre-stressed and the non
redistribution as shown previously due to the increase of the con- pre-stressed wire, the tendency is the same; there is a decrease
tact area that promotes a decrease of the contact pressure. While of the maximum peak value in the first thousands of cycles and
the maximum SWT value is distributed along the entire contact then stabilises. Nevertheless, the influence of the tensile stresses
surface in a fretting cycle in which there is no wear, as the wear oc- produced as a consequence of the bending leads to a drastic in-
curs a peak value appears in the edge of the contact. Moreover as a crease in the peak value, which could reduce considerably the fret-
consequence of the wear and the widening of the wear scar ting induced crack initiation life. Furthermore Fig. 13b shows that
this peak value suffers an outward movement, the same tendency the location of the maximum SWT peak value in the axial direction
presented previously with the axial stress r11 and the shear stress after 50,000 cycles, in both cases, is at the edge of the contact. The
r13. distance from the centre of the wear scar is a little higher in the
On the other hand the 3D SWT contour plots are shown in pre-stressed case than in the non pre-stressed case. So reducing
Fig. 12c for the no wear case and in Fig. 12d after 50,000 fretting the bending of the wire can increase drastically the life of this
wear cycles. The wear produces two symmetrical SWT peaks in component.
both edges of the axial direction. This peak value starts in the Fig. 14 shows the crack initiation locations of the bottom wear
middle of the transversal direction of the wear scar and moves scars after carrying out fretting wear tests with 4 N–65 lm
outwards with decreasing value. Furthermore from both (Fig. 14a) and 2 N–130 lm (Fig. 14b) and 20,000 cycles. Two sym-

(a) 20 N =0
(b) 20
N =5000
15 N =20000 15
N =50000
SWT [MPa]
SWT [MPa]

10 10

5 5

0 0
-0.2 -0.1 0 0.1 0.2 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
4
x coord [mm] Number of Cycles n [10 ]

(c) 20 (d) 10
15 8
SWT [MPa]

SWT [MPa]

6
10
4
5
2
0 0
0.2 0.2
0.1 0.1 0.1 0.1
0 0 0
x coord -0.1 y coord -0.1 0
-0.2 -0.1 -0.2 -0.1
[mm] [mm] x coord y coord
[mm] [mm]

Fig. 12. FE critical plane SWT plots for 1 N–130 lm and 2000 MPa pre-stress: (a) longitudinal SWT value plot for different number of cycles, (b) evolution of SWT peak value,
(c) 3D plot of the SWT value without wear, and (d) 3D SWT value plot after 50,000 cycles.
18 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

(a) 20 (b) 5 0 MPa

0 MPa
2000 Mpa 4 2000 MPa
15

SWT [MPa]
SWT [MPa]
3
10
2

5
1

0 0
0 1 2 3 4 5 -0.2 -0.1 0 0.1 0.2
Number of Cycles n [10 ] 4 x coord [mm]

Fig. 13. FE critical plane SWT plots for 1 N–130 lm with and without pre-stress: (a) SWT peak value evolution and (b) longitudinal plot of the maximum SWT value.

metrical transversal cracks at a specific distance from the centre of

the wear scar are produced. This is in accordance with the SWT
critical plane plots presented previously in Fig. 13b. On the other
hand the crack propagation is produced in the longitudinal plane
to the contact surface, as a consequence of the fibre structure pre-
sented by the wire. This is because during the cold drawn process
the pearlitic structure of the eutectoid steel is formed by lamellas
of cementite and ferrite oriented in the longitudinal direction of
the wire. The wear scars obtained with the FE wear simulation
model, for the two cracks experimental cases presented previously
are shown in Fig. 15. The good correlation of this methodology is
shown in the dimensions of the wear scars.
The life predictions made combining the SWT critical plane ap-
proach and the damage accumulation model, which takes into ac-
count the continuous evolution of the subsurface stresses due to
wear removal, are summarised in Table 7. The life predictions have
been carried out according to the fatigue coefficients obtained from
Fig. 15. FEM simulated wear scars for: (a) 4 N–65 lm–2  104 and (b) 2 N–
the three methodologies, based on the monotonic tension test 130 lm–2  104.
properties of the wire and from the coefficients obtained from
the axial fatigue tests reported by Beretta. In all cases the FE wear decreases drastically. Nevertheless the predicted crack location is
simulation has been carried out for 50,000 cycles, so in the cases closest to that one obtained experimentally. The medians method,
where no crack appears the maximum accumulated damage value which is the last method used to obtain the fatigue coefficients of
and the location of this damage has been listed. The crack location the thin steel wire analysed in this study is the less conservative
value has been considered as the distance from the centre of the method. In the case of 2 N–130 lm no crack is predicted. Moreover
wear scar to the predicted crack position. On the other hand, the the crack location position is over predicted. On the other hand the
experimental tests have been carried out till 200,000 cycles, so fatigue coefficients obtained from the fatigue tests reported by Be-
conditions in which cracks could nucleate at longer number of cy- retta, looks to be the most conservative, under predicting cracks in
cles has not been predicted. all cases. Nevertheless the predicted crack position is the closest to
The Manson’s universal slopes method predict correctly the the one obtained experimentally.
lives in the 130 lm cases; nevertheless under predict cracks in In all cases it is clear that the increase of the load reduce dras-
the case of 2 N–65 lm. The modified Muralidharan’s universal tically the life of the wire, nevertheless, it is not clear that an in-
slopes method is more conservative that the previous method, crease of the stroke also reduce drastically the life of the wire as
mainly with higher loads such us 4 N–65 lm, where the life shown in the experimental results reported in this study and in
180 µm
80 µm

310 µm
265 µm

2x

2x

Fig. 14. Failure of the wire under different conditions fretting wear tests: (a) 4 N–65 lm–2  104 and (b) 2 N–130 lm–2  104.
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 19

Table 7
Comparison of measured and predicted lives and failure locations under different SWT coefficients fitting methods (x represents the location from the centre of the wear scar, D
represent the damage after 50  103 cycles).

Test conditions Experimental Nf Predicted Nf

Universal slopes Modified universal slopes Medians (steels) Beretta
90°–4 N–65 lm (crack) (x = 20–50 lm) 2  104–1  105 5.5  103 (x = 92.9 lm) 200 (x = 0 lm) 1.35  104 (x = 92.9 lm) 500 (x = 18.31 lm)
90°–2 N–65 lm (no crack) 2  105 2.55  104 (x = 80.5 lm) 7.5  103 (x = 80.5 lm) D = 0.2042 (x = 80.5 lm) 2.5  103 (x = 68.13 lm)
90°–2 N–130 lm (crack) (x = 80 lm) 1.5  104–5  104 2.45  104 (x = 142 lm) 1.05  104 (x = 117.5 lm) D = 0.258 (x = 159.7 lm) 3  103 (x = 105.2 lm)
90°–1 N–130 lm (no crack) 2  105 D = 0.313 (x = 142 lm) D = 0.314 (x = 142.3 lm) D = 0.0208 (x = 142.3 lm) 4.5  103 (x = 105.2 lm)

the conclusions reported by Wang et al. [11] where this effect was position (y = 0 coord position) for 4 N–65 lm and 5500, Fig. 16c
associated with an increase of the tangential force. Taking into ac- and d shows the plot for 2 N–130 lm and 24,500 cycles. In both
count this last statement, the values predicted with the proposed cases the model predicts the symmetrical positions of the crack
methodology for 2 N–65 lm and 2 N–130 lm are similar, so there location as shown in the experimental tests. Fig. 16e and f repre-
are more aspects that have to be analysed in order to improve this sents the no crack case prediction plot for the 1 N–130 lm case.
methodology. These aspects will be summarized in the next
section. 4.4. Fretting wear life prediction methodology discussion
Fig. 16 shows the damage location plots of the two crack cases
presented previously, using the fatigue coefficients obtained from This is the first step to developing a consistent methodology for
Manson’s universal slope method, which is the most consistent predicting fretting wear induced cracks in the thin steel wires used
with the experimental lives. While Fig. 16a and b shows the top in wire rope systems. The presented FE fretting wear simulation
view and the axial cut in the transversal maximum damage model with the combined SWT critical plane and the cumulative

(a) (b) 0 1

-0.01
0.8
-0.02
ycoord [mm]
xcoord [mm]

0.6
-0.03
0.4
-0.04
0.2
-0.05

-0.06 0
-0.2 -0.1 0 0.1 0.2
y coord [mm] x coord [mm]

(c) (d) 0 1

-0.01 0.8
-0.02
xcoord [mm]

ycoord [mm]

0.6
-0.03
0.4
-0.04
0.2
-0.05

-0.06 0
-0.2 -0.1 0 0.1 0.2
y coord [mm] x coord [mm]

(e) (f) 0 1

-0.01 0.8

-0.02
ycoord [mm]
xcoord [mm]

0.6
-0.03
0.4
-0.04
0.2
-0.05

-0.06 0
-0.2 -0.1 0 0.1 0.2
y coord [mm] x coord [mm]

Fig. 16. Predicted location of the failure according to Manson’s universal slope coefficients: 4 N–65 lm–5.5  103, (a) top view, (b) longitudinal cut; 2 N–130 lm–2.45  104,
(c) top view, (d) longitudinal cut; 1 N–130 lm–5  104, (e) top view, and (f) longitudinal cut.
20 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21

damage methodology, has demonstrated that there are some as- the crack could initiate at the same time, but due to the effect
pects that have to be improved in order to predict this complex of the wear in the case of 2 N–130 lm the crack could propa-
phenomenon in the thin steel wires. gate and in the case of 2 N–65 lm the crack could arrest and
As reported by Meggiolaro and Castro [33], the different meth- disappear.
ods proposed in this work for obtaining the fatigue coefficients of Future work will be focused on a more exhaustive crack path
the wire from the uniaxial monotonic tests, may result in life pre- identification under different load and stroke conditions, taking
diction errors of an order of magnitude, which is in accordance into account also the fretting fatigue tests proposed by Wang
with the values obtained in Table 7 in specific loading cases and et al. [11] for wider diameters, in which more realistic conditions,
methods (over predicting or under predicting cracks). So they pro- close to those presented in wire ropes could be tested and used for
posed these methods as the first stages of design, requiring the developing a more robust FE wear crack nucleation and propaga-
experimental fatigue parameters for correct predictions. tion methodology for thin steel wires.
On the other hand, when using the fatigue coefficients obtained
from the experimental fatigue data reported by Beretta and Mat-
teazzi [34] for 1.97 mm wire, the results are over conservative. 5. Conclusions
As shown in the S–N curves reported in the literatures [5,8,34], dif-
ferent diameters thin steel wires could have different fatigue prop- This paper has presented a finite element prediction methodol-
erties. This is because the fatigue properties of the wires are related ogy which combines a FE wear model and a critical plane SWT
to different parameters: such as the interlamellar spacing of the damage accumulation approach for the prediction of fretting wear
pearlite obtained by the cold-drawn process, the diameter of the induced cracks in thin steel wires. Four different methods for esti-
wire or the size of the inner defects. The diameter of the wire stud- mating the fatigue coefficients of the steel wire have been analysed
ied in this work is smaller than those ones reported in the litera- and the fatigue life predictions have been compared to the fretting
ture. Moreover this wire presents higher ultimate tensile wear experimental tests carried out on thin steel wires. The main
strength and higher yield strength. So future work will be focused conclusions are:
on characterizing the fatigue behaviour of this wire with the com-
plexity associated with the very small diameter.  A significant reduction of the fretting wear frictionally-induced
The results obtained by the combined FE wear model and SWT multiaxial contact stresses within the first few thousands of
critical plane damage methodology, are mainly over-conservative. cycles is predicted. The maximum tensile stresses are located
One approach used by many researchers to solve this problem is in the edge of the wear scar in the sliding direction.
the application of the volume averaging techniques [15–16], such  The friction coefficient and the bending induced tensile stres-
as these methods can capture in more realistic way the stress field ses in the wire are two of the parameters that can reduce
corresponding to the crack nucleation. On the other hand, to im- crack nucleation life. Low friction coefficients and low bend-
prove the prediction corresponding to the crack location Ding ing pre-stresses are indispensable for increasing the life of
et al. [36] proposed to combine the Ruiz parameters, which empha- the wire.
sise the importance of the frictional work, with the SWT critical  The high effect presented by the bending induced tensile stres-
plane approach. When using these types of parameters with vol- ses combined with the fretting multiaxial stresses leads to
ume averaging of the SWT parameter, as the average grain size, cracks in mode I.
Ding et al. [37] reported the effectiveness in the incorporation  Symmetrical cracks, about the centre of the wear scar, are pre-
the wear effects along with the crack nucleation prediction. dicted, as shown experimentally. However, the locations of the
Moreover in order to simplify this study a mesh size which is cracks obtained experimentally are closer to the centre than the
optimum for the wear simulation as described in [21] was used. predicted ones.
This mesh size gives good contact results in the great part of  The lives predicted with this methodology are conservative rel-
the fretting wear test and smaller mesh sizes could increase dras- ative to those obtained experimentally. Manson’s method and
tically the computational time. Nevertheless it is known that the the median’s method gives lives closer to those obtained exper-
stress and strain values are mesh size dependent. Therefore, it is imentally, while those predicted with the fatigue coefficients
proposed to perform a convergence study with respect to the obtained from the modified universal slopes method and Bere-
stress values and the SWT critical plane damage model. Related ta’s fatigue data are more conservative.
to this it is necessary to take into account the stress gradients  While this approach predicts that increasing load reduces dras-
during the fretting problem in the life prediction. It means that tically the life of the wire, it does not predict that an increasing
volume, line or area averaging techniques have to be taken into stroke also reduces drastically the life of the wire, as shown
account. Around this aspect Sum et al. [17] has previously shown experimentally.
that, in fact, if one averages integration point stresses, or as in  Finally different solutions for the improvement of this method-
this paper uses a centroidal stresses, then the FE essentially pro- ology in future work are proposed. These include: the fatigue
vides a volume averaging mechanism, whereby the element size characterization of the thin steel wires; the introduction of vol-
is the volume or area over which averaging is done. Therefore, ume-averaging techniques based on the element size or the
it could be more important to carry out an analysis of the life Ruiz parameters in the SWT critical plane approach to improve
predication for the different loading cases based on the volume the life predictions and the crack locations; development of
of each element, than performing the convergence study with re- crack nucleation and propagation methodologies which can
spect to the stress and strain values. explain better the effect of crack propagation or arrest due to
One key factor that has to be considered in this methodology is the wear and perform fretting-fatigue tests for a more exhaus-
the introduction of a crack propagation model in combination with tive crack path identification which could help in developing a
the developed crack initiation methodology, such as the methodol- more robust FE wear crack nucleation and propagation method-
ogy proposed by Madge et al. [38] for 2D problems. This approach ology for thin steel wires.
could explain, that, e.g. for the 2 N–65 lm and 2 N–130 lm cases
 A. Cruzado et al. / International Journal of Fatigue 55 (2013) 7–21 21

Acknowledgments [16] Fouvry S, Duo P, Prruchaut Ph. A quantitative approach of Ti–6Al–4V fretting
damage: friction, wear and crack nucleation. Wear 2004;257:916–29.
[17] Sum WS, Williams EJ, Leen SB. Finite element, critical-plane, fatigue life
The authors would like to thank the Basque Government for the prediction of simple and complex contact configurations. Int J Fatigue
financial support under the Universidad Empresa program in the 2005;27:403–16.
[18] Madge JJ, Leen SB, McColl IR, Shipway PH. Contact-evolution based prediction
frame of the SIVICA Project (Ref. UE2009-5). The authors also
of fretting fatigue life: effect of slip amplitude. Wear 2007;262:1159–70.
would like to thank the reviewers for their helpful comments [19] Zhang T, McHugh PE, Leen SB. Computational study on the effect of contact
and discussions. geometry on fretting behaviour. Wear 2011;271:1462–80.
[20] Cruzado A, Hartelt M, Wäsche R, Urchegui MA, Gómez X. Fretting wear of thin
steel wires: Part 1. Influence of contact pressure. Wear 2010;268:1409–16.
[21] Cruzado A, Urchegui MA, Gómez X. Finite element modeling and experimental
References validation of fretting wear scars in thin steel wires. Wear 2012;289:26–38.
[22] McColl IR, Ding J, Leen SB. Finite element simulation and experimental
[1] Schrems KK. Wear related fatigue in a wire rope failure. J Test Eval validation of fretting wear. Wear 2004;256:1114–27.
1994;22:490–9. [23] Fouvry S, Paulin C, Meunier C. Finite element modelling of fretting wear
[2] Waterhouse RB. Fretting in steel ropes and cable – a review. ASTM Spec Tech surface evolution: application to a Ti–6Al–4V contact. Wear 2008;264:26–36.
Publ 2002;1425:3–14. [24] Mary C, Fouvry S. Numerical prediction of fretting contact durability using
[3] Urchegui MA, Tato W, Gómez X. Wear evolution in a stranded rope subjected energy wear approach: optimisation of finite-element model. Wear
to cyclic bending. J Mater Eng Perform 2007;17:550–60. 2007;263:444–50.
[4] Llorca J, Sánchez-Gálvez V. Fatigue limit and fatigue life prediction in high [25] Abaqus User’s Manual, version 6.9. USA: Hibbit, Karlsson & Sorensen, Inc.;
strength cold drawn eutectoid steel wires. Fatigue Fract Eng Mater Struct 2009.
1985;12:31–45. [26] Navarro C, Muñoz S, Domínguez J. On the use of multiaxial fatigue criteria for
[5] Beretta S, Boniardi M. Fatigue strength and surface quality of eutectoid steel fretting fatigue life assessment. Int J Fatigue 2008;30:32–44.
wires. Int J Fatigue 1999;21:329–35. [27] Fridrici V, Fouvry S, Kapsa P, Perruchaut P. Prediction of cracking in Ti–6Al–4V
[6] Petit J, Sarrazin-Baudoux C, Lorenzi F. Fatigue crack propagation in thin wires alloy under fretting-wear: use of the SWT criterion. Wear 2005;259:300–8.
of ultra high strength steels. Proc Eng 2010;2:2317–26. [28] Muñoz S, Proudhon H, Domínguez J, Fouvry S. Prediction of the crack extension
[7] Lambrighs Kasper, Verpoest Ignace, Verlinden Bert, Wevers Martine. Influence under fretting wear loading conditions. Int J Fatigue 2006;28:1769–79.
of the load ratio on the threshold intensity factor range for heavily drawn steel [29] Das J, Sivakumar SM. An evaluation of multiaxial fatigue life assessment
wires. Eng Fail Anal 2011;18:694–9. methods for engineering components. Int J Pres Ves Pip 1999;76:741–6.
[8] Navaei Alvar E, Aghazadeh Mohandesi J. Fatigue damage accumulation in cold- [30] Del Llano-Vizcaya L, Rubio-González C, Mesmacque G, Cervantes-Hernández T.
drawn steel wire under variable loading. Mater Des 2010;31:2018–24. Multiaxial fatigue and failure análisis of helical compression springs. Eng Fail
[9] Brighenti Roberto, Carpinteri Andrea, Vantadori Sabrina. Influence of residual Anal 2006;13:1303–13.
stresses on fatigue crack propagation in pearlitic cold-drawn steel wires. Mater [31] Manson SS. Fatigue: a complex subject-some simple approximations. Exp
Sci Forum 2011;681:229–35. Mech 1965;5:193–226.
[10] Zhang DK, Ge SR, Quiang YH. Research on the fatigue and fracture behaviour [32] Muralidharan U, Manson SS. Modified universal slopes equation for estimation
due to the fretting wear of steel wire in hoisting rope. Wear 2011;271:866–74. of fatigue characteristics. J Eng Mater – T ASME 1988;110:55–8.
[11] Wang DG, Zhang DK, Wang SQ, Ge SR. Finite element analysis of hoisting rope [33] Meggiolaro MA, Castro JTP. Statistical evaluation of strain-life fatigue crack
and fretting wear evolution and fatigue life estimation of steel wires. Eng Fail initiation predictions. Int J Fatigue 2004;26:463–76.
Anal 2013;27:173–93. [34] Beretta S, Matteazzi S. Short crack propagation in eutectoid steel wires. Int J
[12] Wang DG, Zhang DK, Ge SR. Effect of displacement amplitude on fretting Fatigue 1996;18:451–6.
fatigue behavior of hoisting rope wires in low cycle fatigue. Tribol Int [35] Verpoest I, Aernoudt E, Deruyttere A, De Bondt M. The fatigue threshold,
2012;52:178–89. surface condition and fatigue limit of steel wire. Int J Fatigue 1985;7:199–214.
[13] Wang DG, Zhang DK, Ge SR. Finite element analysis of fretting fatigue [36] Ding J, Sum WS, Sabesan R, Leen SB, McColl IR, Williams EJ. Fretting fatigue
behaviour of steel wires and crack initiation characteristics. Eng Fail Anal predictions in a complex couplings. Int J Fatigue 2007;29:1229–44.
2013;28:47–62. [37] Ding J, Houghton D, Williams EJ, Leen SB. Simple parameters to predict effect
[14] Szolwinski MP, Farris TN. Mechanics of fretting fatigue crack formation. Wear of surface damage on fretting fatigue. Int J Fatigue 2011;33:332–42.
1996;198:93–107. [38] Madge JJ, Leen SB, Shipway PH. A combined wear and crack nucleation–
[15] Araújo JA, Nowell D. The effect of rapidly varying contact stress fields on propagation methodology for fretting fatigue prediction. Int J Fatigue
fretting fatigue. Int J Fatigue 2002;24:763–75. 2008;30:1509–28.